B200 FP4: M3 EAGLE vs Off Speculative Decoding
Speculative decoding comparison of M3 EAGLE versus Off on B200 FP4 (NVIDIA Blackwell) running MiniMax M3 428B. Throughput, cost, and interactivity differences across LLM workloads. Use the chart controls below to switch sequences and metrics — same interactions as the main inference chart.
MTP acceptance-rate implementations differ across inference engines. Points from different engines are not directly comparable on the same curve — throughput and cost at matched interactivity may reflect engine-level differences rather than pure speculative decoding gains. Interpret cross-engine comparisons with caution.
Throughput at 71 tok/s/user on MiniMax M3 428B (B200 FP4): M3 EAGLE hits 8846 tok/s/chip, Off hits 8368. Per-million costs land at $0.05 and $0.06 respectively. M3 EAGLE is 6% cheaper per token; M3 EAGLE delivers 6% more tok/s/chip. Speculative decoding trades extra compute on draft tokens for fewer decoding steps — the payoff depends on sequence length and batch size.
Around the middle of the 22–220 tok/s/user interactivity band, at 121 tok/s/user on MiniMax M3 428B (B200 FP4): M3 EAGLE runs 5814 tok/s/chip at $0.08/M tokens, Off runs 4517 at $0.11/M. M3 EAGLE is 29% cheaper per token; M3 EAGLE delivers 29% more tok/s/chip. Gains from speculative decoding vary by workload; short-output prompts tend to benefit less.
At 171 tok/s/user on MiniMax M3 428B (B200 FP4), M3 EAGLE delivers 5229 tok/s/chip at $0.09 per million tokens; Off delivers 1138 tok/s/chip at $0.42. M3 EAGLE is 360% cheaper per token; M3 EAGLE delivers 360% more tok/s/chip. Speculative decoding accepts draft tokens to reduce per-token latency — gains vary by workload and prompt distribution. (Numbers reflect this URL's pinned 8k/1k · fp4 workload — changing sequence or model updates both the table and chart; the table stays pinned to this page's precision, so precision toggles in the controls affect the chart only.)
| Metric | Interactivity (tok/s/user) | Interactivity (tok/s/user) | Interactivity (tok/s/user) |
|---|---|---|---|
| Throughput (tok/s/chip) | M3 EAGLE:8845.7Off:8367.9 | M3 EAGLE:5813.8Off:4517.0 | M3 EAGLE:5229.1Off:1137.8 |
| Cost ($/M tok) | M3 EAGLE:$0.054Off:$0.057 | M3 EAGLE:$0.083Off:$0.106 | M3 EAGLE:$0.092Off:$0.422 |
| tok/s/MW | M3 EAGLE:5172906Off:4893496 | M3 EAGLE:3399910Off:2641510 | M3 EAGLE:3057957Off:665389 |
| Concurrency | M3 EAGLE:~33Off:~232 | M3 EAGLE:~20Off:~64 | M3 EAGLE:~70Off:~3 |
Inference Performance
Agentic inference metrics from the AgentX scenario and fixed-sequence inference metrics across models, hardware configurations, and serving parameters.